Pranjal Sahu

Faced the same issue with ITK ImageGrid. Works fine in new conda environment.

@jhlegarreta Yes, I will do the suggested changes. I am checking if one of the functions can be removed from the codebase.

@jhlegarreta I am adding a commit message regarding the removal of the Search method. That is important and should be present for future reference.

This I think is by design. Similar results will be seen when using **vtkPolyDataConnectivityFilter** for triangle meshes.

@N-Dekker Actually this commit is breaking it [ENH: Make itk::Matrix trivially copyable, following Rule of Zero](https://github.com/InsightSoftwareConsortium/ITK/commit/59ebc33c6e835bd8ed52e54e40ee15a7fa24377e)

I am adding some float combinations in the GTest to see the behavior. Because the python wrappings are failing only for float combinations.

@N-Dekker I am trying your suggested changes.

Simply reverting that commit didn't help. But I found something that might give us come clue. The object type is different for MatrixF33 and (MatrixF44, matrixF22 etc.). Only MatrixF33 is...

I am doing a binary search to see where it fails. I saw that it worked for this commit [N-Dekker/Work-around-GCC4-error-is_trivially_copyable](https://github.com/InsightSoftwareConsortium/ITK/commit/d78555518fe12e543936c6871d8463ff630ec9ef) after reverting [ENH: Make itk::Matrix trivially copyable, following Rule of...

Ok, I got the issue. vnl_matrix_fixed is not wrapped for float combinations ! Checking the build now. [vnl_matrix_fixed.wrap](https://github.com/InsightSoftwareConsortium/ITK/blob/master/Modules/Core/Common/wrapping/vnl_matrix_fixed.wrap#L15 ) Update: It works !